I ran a calibration when someone previously had asked if I did or not and got the same result. Sorry, Thought I stated it but guess I didn't. I wasn't sure how long the scope was on but it wasn't just a few minutes. My scopes now been on a few hours now and I'm running another and will report back shortly.
This is definitely bug territory though. Changing the volts/div on the channel with the probe scales the error on the channel without the probe. Changing the volts/div on the channel without the probe also changes the phantom vrms. Another poster said it showed the error on the channel with no probe while the coupling was set to ground. That should remove any doubt.
Edit: As predicted and expected, same after another calibration.
Edit 2: Another interesting tidbit - if probe is in channel 1, but the only measurement displayed is vrms2, it reads close to 0 as it should. The moment you turn on vrms1 variable, vrms2 jumps up.